of their relatively limited reliability, and random accuracy in time and space,
crowdsourced observations have not been widely integrated in hydrological and/or
hydraulic models for flood forecasting applications. Instead, they have generally
been used to validate model results against observations, in post-event analyses.
Different studies addressed the issue of assimilation of distributed observations in
distributed and semi-distributed hydrological models (e.g. [40–43]). Neither of the
previous studies considers the dynamic nature of data from heterogeneous sensors
which provide an intermittent signal in time and space. In fact, the information
coming from a specific sensor might be sent just once, occasionally or in time steps
that are non-consecutive, i.e. with intermittent observations having different
lifespans.
A number of studies have developed methods for using crowdsourced citizensbased observations in water-related models [44–56]. In particular, crowdsourced
information are used for directly creating deterministic or probabilistic flood maps
[48], derive stream discharges and flow velocities fields [57] and flood extent [52]. In
alternative, crowdsourced data have been used for validating flood models
[44, 56]. A detailed review on the use of citizen observations for flood modelling
applications is provided in Assumpção et al. [58]. However, none of the previous
studies assessed the usefulness of citizen observations for improving flood predictions [39, 59]. The first attempts to study the effects of assimilating crowdsourced
citizen observations in hydrological and hydraulic models for improving flood
prediction in real-time applications are reported in Mazzoleni et al. [60–62] and
Mazzoleni [63]. Just recently, Mazzoleni et al. [64] proposed two innovative
approaches to assimilated qualitative flow data within hydrologic routing models.
In this chapter, we describe the proposed innovative methods to assimilate
heterogeneous intermittent observations, coming from social sensors, within hydrological and hydrodynamic modelling to improve flood prediction. This research was
carried out under the framework of the European project WeSenseIt (https://www.
wesenseit.com/) [65].
2 Crowdsourced Observations
In this chapter, we consider two different types of sensors to measure hydrological
variables such as water level: static physical (StPh) and static social (StSc) sensors
(see Fig. 1). In addition, also dynamic social sensors may be used but are not
included in this chapter. An example of a static social sensor is a staff gauge located
in a strategic point of the river used by citizens to estimate water depth values using a
mobile phone app to send CS observations using the QR code as geographical
reference point. An example of dynamic sensor is a mobile app allowing any citizen
to send the information related to the distance between the water profile and the river
bank using a mobile app at random locations along the river. It might be in fact
difficult to estimate the water depth value without having any indication about river
212
M. Mazzoleni et al.
crowdsourced observations have not been widely integrated in hydrological and/or
hydraulic models for flood forecasting applications. Instead, they have generally
been used to validate model results against observations, in post-event analyses.
Different studies addressed the issue of assimilation of distributed observations in
distributed and semi-distributed hydrological models (e.g. [40–43]). Neither of the
previous studies considers the dynamic nature of data from heterogeneous sensors
which provide an intermittent signal in time and space. In fact, the information
coming from a specific sensor might be sent just once, occasionally or in time steps
that are non-consecutive, i.e. with intermittent observations having different
lifespans.
A number of studies have developed methods for using crowdsourced citizensbased observations in water-related models [44–56]. In particular, crowdsourced
information are used for directly creating deterministic or probabilistic flood maps
[48], derive stream discharges and flow velocities fields [57] and flood extent [52]. In
alternative, crowdsourced data have been used for validating flood models
[44, 56]. A detailed review on the use of citizen observations for flood modelling
applications is provided in Assumpção et al. [58]. However, none of the previous
studies assessed the usefulness of citizen observations for improving flood predictions [39, 59]. The first attempts to study the effects of assimilating crowdsourced
citizen observations in hydrological and hydraulic models for improving flood
prediction in real-time applications are reported in Mazzoleni et al. [60–62] and
Mazzoleni [63]. Just recently, Mazzoleni et al. [64] proposed two innovative
approaches to assimilated qualitative flow data within hydrologic routing models.
In this chapter, we describe the proposed innovative methods to assimilate
heterogeneous intermittent observations, coming from social sensors, within hydrological and hydrodynamic modelling to improve flood prediction. This research was
carried out under the framework of the European project WeSenseIt (https://www.
wesenseit.com/) [65].
2 Crowdsourced Observations
In this chapter, we consider two different types of sensors to measure hydrological
variables such as water level: static physical (StPh) and static social (StSc) sensors
(see Fig. 1). In addition, also dynamic social sensors may be used but are not
included in this chapter. An example of a static social sensor is a staff gauge located
in a strategic point of the river used by citizens to estimate water depth values using a
mobile phone app to send CS observations using the QR code as geographical
reference point. An example of dynamic sensor is a mobile app allowing any citizen
to send the information related to the distance between the water profile and the river
bank using a mobile app at random locations along the river. It might be in fact
difficult to estimate the water depth value without having any indication about river
212
M. Mazzoleni et al.